Introduction

Ultrasa is created using RASA X. Basically, it is a chatbot that can chitchat with users.You might wonder how Ultrasa works. Here’s the answer. Link to GitHub Repository

For each incoming message, Ultrasa will predict intent of the message and give response based on intent. If the related intent requires some actions like action_suggestplaylist, Ultrasa will carry out the action and then give final output.

Components Used

To predict intent, We used a sequence of components.

• Pre-trained Model (BERT)
• Tokeizer (LanguageModelTokenizer)
• Featurizer (LanguageModelFeaturizer)
• Classifier (DIETClassifier)

Intents

By taking reference from Google’s GoEmotions, Ultrasa is able to identify 27 types of emotions.

Chatbot Functions

Our Ultrasa has equipped will the below functions:

1. Funny Quote Telling
   • Action will be triggered when the emotion of user is identified as “Embarrassment”.
   • The quote are randomly picked from https://www.keepinspiring.me/funny-quotes/

     

2. Hot Song Introduction
   • When Ultrasa is asked to introduce a song, it will randomly return a hot song from BillBoard Hot 100.

     

3. Music Playlist Introduction
   • The intent of each message are stored in SlotSet in wach conversation in actions.py.
   • When Ultrasa is asked to introduce a playlist, it will randomly return a youtube music playlist url based on user’s mood.

     

4. Food Recipe Recommendation
   • Ultra will recommend recipe based on user’s target calories
   • It will then provide cooking time and cooking instructions

     

Deployment

Ultrasa has been successfully deployed to website and Telegram.

Link to GitHub Repository